In November 2022, the San Francisco-based OpenAI released ChatGPT onto an unsuspecting world and caused a sensation. The AI chatbot was able to generate lengthy text responses to complicated questions. It was a personal triumph for Sam Altman, OpenAI’s chief executive and co-founder. But a year later, the business world was stunned when Altman was ousted and then reinstated as CEO. The story of that brief coup reveals a company riven with doubt and suspicion, wrangling with the great responsibility of how to shape a world-changing technology.
Sam Altman once promised the world that he was no Mark Zuckerberg. “I don’t have super-voting shares. I don’t want them,” he said in June 2023. “The board can fire me. I think that’s important.” But behind the scenes, the non-profit board of OpenAI, with a majority of supposedly “independent” directors, was finding, to its growing frustration, that Altman really called the shots.
In the autumn of 2022, following ChatGPT’s spectacular release, the board was divided over the appointment of a new director to advise on AI safety. “There was a little bit of a power struggle,” said Brian Chesky, the Airbnb chief executive and Altman ally who was a prospective board member.
The situation grew worse after three board members in the pro-Altman camp stepped down in quick succession. One remaining member, Adam D’Angelo, the Quora chief executive, said he wanted to improve OpenAI’s corporate governance. These concerns became more urgent as they realised how rapidly its decisions were turning into potentially grave ones.
“The stakes of the company are ramping up over time,” said Helen Toner, a board member.
There were also signs that some directors were losing trust in Altman. They were concerned that he was not taking the company’s deployment safety board (DSB) seriously. The DSB was intended to review new products for risks.
In some instances, OpenAI’s partner and investor, Microsoft, fell foul of the DSB. The OpenAI board was alarmed when it was informed about such a setback by an employee — who stopped a board member in the hallway — rather than Altman himself.
Sam Altman with the Microsoft chief Satya Nadella, right, in San Francisco on November 6, 2023. Altman would be gone, briefly, later that month
JUSTIN SULLIVAN/GETTY IMAGES
• Watchdog drops investigation into Microsoft’s $13bn stake in OpenAI
Then, one night in the summer of 2023, a board member overheard a person at a dinner party discussing how inappropriate it was that returns from OpenAI’s Startup Fund were not going to the organisation’s investors. This was also news to the board. OpenAI did launch a start-up fund in 2021, which it said at the time would be “managed” by OpenAI.
The board began to ask Altman about it and, over the course of months, gradually learnt that he owned the fund personally. This blindsided the independent board members, who felt they should have been informed ahead of time.
In September 2023, Ilya Sutskever, an OpenAI board member and its chief scientist, met with Toner. In a cryptic conversation, he told her: “You should talk to Mira more.”
Mira Murati, then 34, was chief technology officer of OpenAI and in effect ran the company. People who worked with her described her as having immense emotional intelligence and almost no ego. “She’s a calming influence among the founders,” said one former employee.
Murati could also go before audiences and speak authoritatively about the cosmic meaning of OpenAI’s work. She visited The Wall Street Journal in March 2023 for a meeting and showed up wearing black trousers, black high-heeled boots and a black leather jacket, like a cyborg supermodel from the future. She answered questions with an engineer’s unassuming directness.
Altman at a conference in Tel Aviv in June 2023 with Ilya Sutskever. Below, Mira Murati, chief technology officer of OpenAI from May 2022 to September 2024
JACK GUEZ/AFP
Toner called Murati, explaining hat she’d had a confusing conversation with Sutskever and wondered: is something going on? Murati described how what she saw as Altman’s toxic style of running the company had been causing problems for years. In her experience, Altman had a simple playbook: first, say whatever he needed to say to get you to do what he wanted; and second, if that didn’t work, undermine you or destroy your credibility.
Murati had given Altman feedback on all these points in the past, she said. Toner then went back to Sutskever to try to get more information. While he didn’t quite come out and say it, what she took away from their conversation was that he thought Altman should be fired. Sutskever felt Altman was dishonest and causing chaos — which would be a problem for any chief executive, but especially for one in charge of such potentially civilisation-altering technology.
Toner began to talk to the other board members. Not long after, Altman was caught in a lie that ended up being the nail in his proverbial coffin. Sutskever emailed Toner, D’Angelo and a third board member, Tasha McCauley, with a document on Altman that gave dozens of examples of lying or other toxic behaviour. These were largely backed up by screenshots from Murati’s Slack channel. In one of them, Altman tells Murati that the company’s legal department had said a version of GPT didn’t need to go through DSB review. When Murati checked with the company’s top lawyer, he said he could not imagine how Altman would have got that impression; of course it had to go through safety checks.
As they weighed the evidence before them, the board members considered that OpenAI was one of the most promising tech companies on the planet. Surely, they reasoned, they could have their pick of CEOs.
So, on the afternoon of Thursday, November 16, 2023, the three independent board members and Sutskever voted to fire Altman. They also voted to remove Greg Brockman, OpenAI’s president, another Altman ally and board member, who was perceived to be a destabilising influence.
Greg Brockman with OpenAI team members after his and Altman’s return to the company. He had originally been seen by the board as a destabilising influence
GREG BROCKMAN
Murati was at a conference when she received a call asking her to step in as interim chief executive. She agreed to do what she could to stabilise the company. When she asked why they were firing Altman, the board refused to share details.
The next day, Altman was in Las Vegas for a Formula One race when he clicked a Google Meet link for a meeting with Sutskever. He was surprised when he saw the faces of D’Angelo, Toner and McCauley appear on the screen as well — but, ominously, not Brockman, who had been removed from the board minutes before. Sutskever read Altman a short script, saying he was being fired, but gave no specifics about why. In shock, Altman closed the call with: “How can I help?” They told him he could assist with a smooth handover.
Shortly after, Altman was locked out of his computer. The first few moments were pure incredulity. Then the anger came. Minutes later, the news went public via a terse blog post on the OpenAI website that simply said “he was not consistently candid in his communications with the board”.
The news of Altman’s firing instantly became the top story around the world. Altman’s iPhone was getting so many text messages per second that his iMessage app crashed.
At 2pm that Friday, Murati and Sutskever led an all-hands meeting for OpenAI employees, who peppered them for 45 minutes with questions that were all some version of: “What did Sam do?” When one employee asked if they would ever find out, Sutskever replied: “No.”
Altman had been in the final stages of raising investor money to buy employee shares at a valuation of nearly $90 billion, giving many OpenAI employees the chance to become fabulously wealthy. Most staff believed that tender was now as good as dead.
Altman on a visit to London. The reason for his ousting had been that he was “not consistently candid”
RICHARD POHLE FOR THE TIMES
The board members agreed to a virtual call with the executive team that evening. When they logged on, they entered a virtual room full of raw panic. OpenAI’s top lawyer, Jason Kwon, argued that they owed more of an explanation than Altman was “not consistently candid”. By being vague, the board had invited regulatory scrutiny of OpenAI and pain for its employees.
As Friday evening wore on, OpenAI’s board held a series of increasingly contentious meetings. At one point, the executive team, including Murati, gave the board a 30-minute deadline to either explain why they had fired Altman or resign.
Upon legal advice, the board members felt they could not divulge that it had been Murati who had given them some of the most detailed evidence of Altman’s management failings. Yet they felt betrayed and confused that she had seemingly turned on them.
In reality, the board had mistaken Murati’s agreement to help stabilise the company for agreement with their decision to fire Altman. For her part, Murati was shocked that her feedback — which she viewed as transparent, constructive criticism that she had also shared with Altman — would lead the board to assume she supported their actions.
Over the course of the evening, a narrative spread among Altman’s allies that the whole thing was an “Ilya coup”. Sutskever was astounded. He had expected the employees of OpenAI to be relieved to be free from what he saw as Altman’s gaslighting.
On Saturday morning, Altman received a call from D’Angelo and McCauley, who wanted to open a line of dialogue. Over the course of the day, OpenAI employees, including Murati, gathered at his sprawling six-bedroom home, plotting how to bring him back to the company.
By Sunday, Altman and the OpenAI executive team were confident he would be coming back. Murati invited him to the office, without telling the board. Altman flashed his guest pass to his followers on X, writing: “First and last time I ever wear one of these.” But his desire that both he and Brockman would return to the board was a sticking point.
The former US Treasury secretary Larry Summers was suggested as a new OpenAI board member
MANDEL NGAN/AFP
The talks dragged on late into the night. Suddenly the board informed Murati that it was naming a new CEO: Emmett Shear, co-founder of the video platform Twitch and a voice for the careful development of AI. Altman and the OpenAI executive team were stunned. They felt the board had not been negotiating in good faith.
A Slack message went out to all employees announcing an all-hands meeting with Shear. Employees responded with middle-finger emojis.
• No OpenAI without Microsoft, says boss after coup
That night, the Microsoft boss Satya Nadella tweeted that he was hiring Altman and Brockman, and invited OpenAI’s other employees to join them. Overnight, more than 700 of the company’s 770 employees signed a blistering letter threatening to quit and go to Microsoft, unless Altman and Brockman were reinstated.
Among the signatures was Sutskever’s. “I deeply regret my participation in the board’s actions,” he wrote on X. “I never intended to harm OpenAI.”
• Co-founder Ilya Sutskever leaves OpenAI
To bring the matter to a close, D’Angelo suggested the former US Treasury secretary Lawrence Summers as a new board member. Altman agreed and gave up his demand for a board seat. By Tuesday, the deal was done: Altman would return to OpenAI as chief executive but not to the board, and there would be an independent investigation into the circumstances of his firing. Sutskever, Toner and Macauley would step down from the board.
Over the course of five sleepless days, Altman had un-fired himself. He was now more powerful than ever. OpenAI employees would come to refer to the whole episode as “the blip”.
This is an edited excerpt from The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future by Keach Hagey (Norton, May 20, £25)